|Home | About | Journals | Submit | Contact Us | Français|
Objective To determine the relation between overweight and obesity in mothers and preterm birth and low birth weight in singleton pregnancies in developed and developing countries.
Design Systematic review and meta-analyses.
Data sources Medline and Embase from their inceptions, and reference lists of identified articles.
Study selection Studies including a reference group of women with normal body mass index that assessed the effect of overweight and obesity on two primary outcomes: preterm birth (before 37 weeks) and low birth weight (<2500 g).
Data extraction Two assessors independently reviewed titles, abstracts, and full articles, extracted data using a piloted data collection form, and assessed quality.
Data synthesis 84 studies (64 cohort and 20 case-control) were included, totalling 1095834 women. Although the overall risk of preterm birth was similar in overweight and obese women and women of normal weight, the risk of induced preterm birth was increased in overweight and obese women (relative risk 1.30, 95% confidence interval 1.23 to 1.37). Although overall the risk of having an infant of low birth weight was decreased in overweight and obese women (0.84, 0.75 to 0.95), the decrease was greater in developing countries than in developed countries (0.58, 0.47 to 0.71 v 0.90, 0.79 to 1.01). After accounting for publication bias, the apparent protective effect of overweight and obesity on low birth weight disappeared with the addition of imputed “missing” studies (0.95, 0.85 to 1.07), whereas the risk of preterm birth appeared significantly higher in overweight and obese women (1.24, 1.13 to 1.37).
Conclusions Overweight and obese women have increased risks of preterm birth and induced preterm birth and, after accounting for publication bias, appeared to have increased risks of preterm birth overall. The beneficial effects of maternal overweight and obesity on low birth weight were greater in developing countries and disappeared after accounting for publication bias.
The continuum of overweight and obesity is now the most common complication of pregnancy in many developed and some developing countries. In the United Kingdom, 33% of pregnant women are overweight or obese.1 In the United States, 12%2 to 38%3 of pregnant women are overweight and 11%4 to 40%3 are obese. In India, 8% of pregnant women are obese and 26% are overweight5 and in China, 16% are overweight or obese.6
Preterm birth is the leading cause of neonatal mortality and morbidity and childhood morbidity7 followed by low birth weight.8 Whether maternal overweight and obesity is associated with increased,9 decreased,10 or neutral risks11 of preterm birth has been debated in the literature, with the uncertainty reflected in the American College of Obstetrics and Gynecology Committee opinion on obesity in pregnancy.12 Even low birth weight, which is typically thought to be reduced in infants of overweight and obese women,3 is sometimes associated with neutral risks.5 To accurately risk stratify a pregnancy at the first antenatal visit, as is standard, it is important to know the effect of overweight and obesity in mothers on preterm birth and low birth weight. We therefore undertook a systematic, comprehensive, and unbiased accumulation and summary of the available evidence from all study designs with a reference group of normal weight women to determine the direction and magnitude of the association of maternal overweight and obesity with preterm birth and low birth weight in singleton pregnancies in developed and developing countries.
We carried out a systematic review and meta-analyses in accordance with the Meta-analysis Of Observational Studies in Epidemiology consensus statement.13
With the help of a librarian we searched Medline (1950 to 2 January 2009) and Embase (1980 to 2 January 2009), using individual comprehensive search strategies. This study was part of a constellation of systematic reviews examining maternal anthropometry and preterm birth and low birth weight (see search strategy in web extra appendix 1). Additional eligible studies were sought by reviewing the reference lists of identified articles.
For the constellation of systematic reviews examining maternal anthropometry, we included randomised trials, cohort studies, and case-control studies if one or more of the following maternal anthropometry variables was assessed as an exposure variable: body mass index (*=assessed before pregnancy, during pregnancy or postpartum), weight*, gestational weight gain, attained weight, or height*; and one or more of the following outcomes was assessed: preterm birth (<37 weeks, 32-36 weeks, and <32 weeks) and low birth weight (<2500 g), very low birth weight (<1500 g), and extremely low birth weight (<1000 g). Studies were restricted to those in English. For this particular systematic review of maternal overweight and obesity, we included studies with any body mass index definition of overweight and obese or very obese, whether from self report, objective measurement, medical charts, or databases.
We excluded duplicate publications, studies published only as abstracts, those involving fewer than 10 patients, and those that examined outcomes in multiples unless stratification was done for singleton versus twin outcomes.
Our primary outcomes were preterm birth (before 37 weeks) and low birth weight (<2500 g) in singletons. Where possible we subdivided preterm birth into spontaneous and induced. Secondary outcomes were late preterm birth (32-36 weeks) and moderate preterm birth (before 32 weeks), and very low birth weight (<1500 g) and extremely low birth weight (<1000 g).
We also reported the following outcomes for studies that met the above inclusion criteria and mentioned intrauterine growth restriction (defined as birth weight <10% for gestational age), birth weight (grams), and gestational age at birth (weeks).
Two assessors (two of ZH, SDM, and SM) independently reviewed titles and abstracts of all identified citations. The full text article was retrieved if either reviewer considered the citation potentially relevant. Two reviewers (two of ZH, SDM and SM) independently evaluated each full text article. Disagreements were settled by discussion and consensus, with a third person as an adjudicator.
From full text articles and using a piloted data collection form, two reviewers independently extracted data on country of origin, years of study, study design, characteristics of participants, outcomes, and information on bias. We included information available from the publications. Inconsistencies were checked and resolved through the consensus process.
We used Review Manager, version 5.0 (Cochrane Collaboration), for statistical analyses. For cohort studies we used relative risks to meta-analyse crude and separately, adjusted, dichotomous data, whereas for case-control studies we used odds ratios to pool crude and separately, matched or adjusted dichotomous data. Continuous data were analysed using a mean difference. Weighting of the studies in the meta-analyses was calculated on the basis of the inverse variance of the study. The random effects model was chosen because it accounts for both random variability and the variability in effects among the studies as we expected a degree of clinical and statistical heterogeneity among the studies, which were all observational. Crude, matched, and adjusted data were initially pooled separately and then matched or adjusted data were pooled together. Where required and when the incidence of the outcome was rare, to be able to pool data, adjusted relative risks were calculated from adjusted odds ratios.14 As is typical in meta-analyses, we did not adjust for multiple analyses. We focused on the combined results of overweight, obese, and very obese; however, where possible we also separately reported results for each individually in the summary tables. Clinical heterogeneity was evaluated. We calculated the I2 value to measure heterogeneity. An I2 value represents the percentage of total variation across studies due to heterogeneity rather than due to chance.15 Values of 25%, 50%, and 75% have been regarded as representing low, moderate, and high heterogeneity.15
Sensitivity analyses were planned a priori using a few chosen groups to examine the effects of level of material wellbeing (developed v developing countries16), study quality (see web extra appendix 2), youth (adolescence v adulthood), and race. Three post hoc sensitivity analyses were carried out (see web extra appendix 3) to examine the effects of self reported compared with measured body mass index; body mass index assessed before pregnancy, during pregnancy, or post partum; and using exact cut-offs for body mass index with a reference body mass index of 20-25 versus those with cut-offs close to this.
Two reviewers (two of ZH, SDM, and SM) independently assessed study quality using a predefined evaluation of six types of biases: selection, exposure, outcome, confounding, analytical, and attrition (see web extra appendix 2). This bias assessment tool has been described in other reviews undertaken by our group on determinants of preterm birth and low birth weight.17
To deal with publication bias we showed results without imputation as well as with imputation: the latter using Duval and Tweedie’s trim and fill method for estimating and adjusting for the number and outcomes of missing studies in a meta-analysis18 19—that is, to adjust for any observed publication bias. A priori we decided to carry out the trim and fill analyses for outcomes with at least 10 studies as there were concerns of reliability for outcomes with fewer studies. We used the generic inverse variance method to calculate study specific weights. These analyses were done using the R statistical and programming software, version 2.9.0. (R Foundation for Statistical Computing, Vienna, Austria).
Overall, 6283 non-duplicated titles and abstracts were identified (fig 11).). After the screening process, 503 citations were selected to undergo review of the full text article, and a further 52 articles were identified from reference lists, yielding a total of 555 full text articles for review. The most common reasons for exclusion were failure to report outcomes of interest and study design.
Eighty four studies were included: 64 cohort studies2 3 4 5 6 9 10 11 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 (58 with pooled data) and 20 case-control studies76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 (19 with pooled data), totalling at least (some studies did not report the number of patients) 1095834 women (fig 1, tables 11 and 22).). The studies originated predominantly from developed countries, although developing countries were also represented. The majority of the studies assessed body mass index by self report. Most studies did not report the timing of body mass index assessment, although when reported it was most commonly at the first antenatal visit.
In the pooled cohort studies the overall risk of preterm birth before 37 weeks did not differ significantly among overweight or obese women with singleton pregnancies (relative risk 1.06, 0.87 to 1.30, 38 studies, fig 22)) compared with women of normal weight (table 33).). However, among overweight and obese women the risk of induced preterm birth was increased (1.30, 1.23 to 1.37, five studies, fig 33).). The heavier the woman, the higher the risk of induced preterm birth before 37 weeks, with overweight, obese, and very obese women having a relative risk of 1.15 (1.04 to 1.27), 1.56 (1.42 to 1.71), and 1.71 (1.50 to 1.94), respectively. The risk of spontaneous preterm birth did not differ (0.93, 0.85 to 1.01, 15 studies). Heterogeneity ranged from 0 to 99%, with most studies in the moderate to high range.
Overweight and obese women had an increased risk of preterm birth before 33 weeks (crude relative risk 1.26, 95% confidence interval 1.14 to 1.39). The heavier the woman, the higher the risk of early preterm birth, with overweight, obese, and very obese women having a relative risk of 1.16 (1.05 to 1.29), 1.45 (1.23 to 1.71), and 1.82 (1.48 to 2.24), respectively.
Compared with the number of studies that presented crude data, few presented matched or adjusted data (table 3). The pooled risks from adjusted or matched data were generally similar in magnitude and direction to that of the pooled crude data—for example, the risk of preterm birth before 37 weeks in overweight or obese women remained non-significant (1.02, 0.68 to 1.54), although the adjusted or matched risk for several outcomes with only one study differed (for example, the adjusted relative risk of spontaneous preterm birth before 37 weeks was 2.29, 95% confidence interval 1.20 to 4.38).
The results of six cohort studies4 25 45 61 72 73 not included in the meta-analysis (the format of the data did not permit pooling) generally supported the pooled data. One study showed an increased risk of preterm birth before 37 weeks45 in overweight and obese women and another showed a slight decreased risk.4 Similar to the pooled data, there were decreases in spontaneous preterm birth before 37 weeks4 72 and increases in the risk of induced preterm birth before 37 weeks.4 54 Preterm birth (32-36 weeks) was significantly increased in overweight and obese women in one study25 but not in another.61 Unlike the pooled data there was no significant increase in preterm birth before 32 weeks.4 25 61
Data from seven case-control studies that examined maternal body mass index as a continuous variable also generally supported the findings of the cohort data. The mean body mass index of women with preterm birth before 37 weeks overall did not differ significantly from those with term births (−0.33 body mass index unit, −1.19 to 0.53), although women with spontaneous preterm birth had a slightly lower body mass index (−0.90, −1.77 to −0.02; table 44).
A few case-control studies reported body mass index as a dichotomous variable (high versus reference; table 55)) There was a trend towards preterm birth before 37 weeks in overweight or obese women overall (crude odds ratio 1.16, 95% confidence interval 0.99 to 1.37), although not in the matched data (odds ratio 1.08, 0.39 to 2.95). The risk of spontaneous preterm birth in overweight or obese women was increased in those in the matched data (1.79, 1.73 to 2.84) but not the crude data (1.00, 0.18 to 5.53). One case-control study that could not be pooled found a trend towards decreased spontaneous preterm birth (crude odds ratio 0.58, 95% confidence interval 0.33 to 1.03).
In the pooled cohort studies, overweight and obese women had a decreased risk of having an infant of low birth weight (relative risk 0.84, 95% confidence interval 0.75 to 0.95, 28 studies, fig 44)) but an increased risk of having an infant of very low birth weight (<1500 g, 1.61, 1.42 to 1.82, two studies) or extremely low birth weight (<1000 g, 1.31, 1.08 to 1.59, one study; table 66).). The heavier the woman, the higher the risk of having an extremely low birth weight infant, with relative risks in overweight, obese, and very obese women of 1.18 (0.94 to 1.47), 1.43 (1.05 to 1.95), and 1.98 (1.36 to 2.89), respectively.
Two cohort studies with non-pooled data showed similar risks of low birth weight in overweight and obese women compared with women of normal weight (adjusted odds ratios 1.4, 95% confidence interval 0.9 to 2.145 and 0.3, 0.1 to 1.0).73
In the seven pooled case-control studies women with low birth weight singletons had a lower maternal body mass index than women with singletons of appropriate weight in both the crude data (−1.15 body mass index units, 95% confidence interval −1.87 to −0.44) and the single study of matched data (−1.20, −1.85 to −0.55; table 4). The single case-control study that dichotomised body mass index into high versus reference also found a decreased risk of infants with low birth weight among mothers with a high body mass index (odds ratio 0.51, 95% confidence interval 0.36 to 0.74; table 5).
In the pooled cohort studies, overweight and obese women had a lower risk of infants with intrauterine growth restriction than women of normal weight (crude relative risk 0.79, 0.72 to 0.88, table 6), and infants with higher mean birth weights by 70.8 g (54.4 g to 87.2 g) despite shorter mean gestations (by −0.06 weeks, 95% confidence interval −0.12 weeks to −0.01 weeks).
One case-control study reported that women with singletons showing intrauterine growth restriction had a lower mean body mass index than women with infants of normal growth (−1.70 body mass index units, 95% confidence interval −2.64 to −0.76; table 4).
Many of the categories in the sensitivity analyses had few studies, limiting our power to draw conclusions. In developing countries, the risk of preterm birth in overweight and obese women were similar to those of women in developed countries (relative risk 0.83, 95% confidence interval 0.61 to 1.12 and 1.09, 0.87 to 1.36; table 77).
No studies were of low quality. There was no significant increase in preterm birth among adolescents compared with adults (0.98, 0.76 to 1.28, one study, and 1.09, 0.95 to 1.25, four studies). Only one study reported on ethnicity; the risk of preterm birth was not significantly increased in overweight and obese black women (0.84, 0.69 to 1.03) or white women (1.03, 0.77 to 1.38).
The decreased risk of low birth weight in overweight and obese women compared with women of normal weight in developing countries was greater than in developed countries (0.58, 0.47 to 0.71, 11 studies v 0.90, 0.79 to 1.01, 20 studies; table 88).). In developing countries, the heavier the woman the smaller the risk of having an infant of low birth weight: relative risks for overweight, obese, and very obese women were, respectively, 0.88 (0.64 to 1.23), 0.39 (0.11 to 1.34), and 0.29 (0.10 to 0.89).
Only one study was of low quality, limiting conclusions on the effect of study quality. Overweight and obese adolescents but not adults were at a decreased risk of having an infant of low birth weight (0.76, 0.63 to 0.92 v 1.0.8, 0.82 to 1.42).
No studies specified whether their population was white and therefore the effect of ethnicity on low birth weight could not be examined.
Quality assessment (tables 99 and 1010)) was based on the evaluation of six types of bias. Selection bias was unlikely as women with high and normal body mass indices were usually drawn from the same populations, whereas exposure bias was possible given that weight was self reported in most studies.
Little bias was present in our outcomes as they had standard definitions and were objectively measured—for example, low birth weight was always defined as birth weight <2500 g.
Confounding variables that might explain part or all of the relation between overweight and obesity and preterm birth and low birth weight were incompletely dealt with in several ways: by exclusion, by matching, by comparison of some variables and determining that they were not significantly different between the exposed and unexposed women, and by using multiple regression to control for some variables that were significantly different between the two groups. Most studies assessed some confounding variables, but none addressed all. Many studies did not calculate a sample size or power calculation. Attrition bias was rare given that follow-up occurred during the hospital admission for birth.
The trim and fill analysis of preterm birth before 37 weeks suggested that nine studies were “missing” from the initially meta-analysed relative risk of 1.06 (95% confidence interval 0.87 to 1.30); when the nine studies were imputed yielding a risk based on a total of 49 studies, the risk of preterm birth before 37 weeks was significantly higher in overweight and obese women than normal weight women (1.24, 1.13 to 1.37, see web extra appendix 4). The trim and fill analysis resulted in no additional imputed studies for preterm birth before 32 weeks (with the original studies showing an increased risk in overweight or obese mothers). The risk of spontaneous preterm birth in overweight or obese women was similar with four additional imputed studies (0.89, 0.81 to 0.97). After accounting for publication bias, the apparent protective effect of overweight or obesity on low birth weight disappeared with the addition of nine imputed studies, yielding an overall risk based on 40 studies (0.95, 0.85 to 1.07, see web extra appendix 4).
In this systematic review and meta-analyses, we determined that overweight and obese women have an increased risk of a preterm birth before 32 weeks, induced preterm birth before 37 weeks, and, accounting for publication bias, preterm birth before 37 weeks overall. The beneficial effects of overweight or obesity on low birth weight were greater in developing countries than developed countries and disappeared after accounting for publication bias.
This systematic review tackles the uncertainty reflected in guidelines from both the American College of Obstetrics and Gynecology and the Institutes of Medicine96 97 on the relation between overweight and obesity in mothers and preterm birth. The 1990 Institutes of Medicine guidelines focused predominantly on problems with birth weight because of the ease of measurement and acknowledged a dearth of information on obese women in particular and on preterm birth in general,96 the leading cause of neonatal morbidity and mortality.7 The revised 2009 guidelines stated that compared with low birth weight, the literature on preterm birth is “more ambiguous because of a less extensive body of epidemiologic evidence”97; however, we included 40 studies on preterm birth. Overweight and obesity were associated with increased risks of both induced preterm birth before 37 weeks and overall preterm birth before 32 weeks, and potentially preterm birth before 37 weeks overall. The significant increase in induced preterm birth in overweight and obese women may account for the trend towards a decrease in spontaneous preterm birth.
To our knowledge this is the first comprehensive systematic review on the effect of maternal overweight or obesity on preterm birth and low birth weight. Two previous studies have tackled a limited portion of the literature. A systematic review on spontaneous preterm birth found no association with maternal anthropometry (likelihood ratio 0.96, 95% confidence interval 0.66 to 1.40).98 However, the quality assessment of studies was limited and several large studies have been published since the literature search ended in 2002. A World Health Organization study meta-analysed 25 datasets identified by researchers attending a 1990 conference but lacked the literature search that is the standard basis of a systematic review.99 Compared with women with higher body mass indices (>75% quartile), women in the lower fourth (<25%) had an increased risk of low birth weight (odds ratio 1.8, 95% confidence interval 1.7 to 2.0) and preterm birth (1.3, 1.1 to 1.4).
The strengths of our meta-analysis include the thoroughness with which the outcomes of preterm birth and low birth weight were assessed (preterm birth was examined before 37 weeks, 32-36 weeks, and before 32 weeks, overall as well as spontaneous and induced, and besides low birth weight we examined very low birth weight and extremely low birth weight). We explored the effect of gradations in maternal body mass index (overweight, obese, and very obese), carried out an extensive quality assessment, and investigated heterogeneity with sensitivity analyses. We compared the results of crude, and matched or adjusted, data to try to determine if the observed perinatal risks were due to body mass index independently or were explained by confounding factors. Finally, we robustly assessed bias using the trim and fill method.
Limitations of this systematic review include potential residual confounding by factors that might account for the observed association between obesity and perinatal outcomes, which were not adjusted for in some or all of the original studies, such as smoking or low socioeconomic status. Gestational weight gain, which was not taken into account by most of the studies, can influence outcomes such as preterm birth and low birth weight. However, prepregnancy body mass index is the strongest predictor of outcomes, not gestational weight gain.100 Moreover, it is useful to be able to predict a woman’s risk of preterm birth or having an infant of low birth weight on the basis of information available at the start of the pregnancy such as prepregnancy body mass index.
We pooled data based on the original studies’ definitions of overweight, obese, and very obese, as have other meta-analyses.101 This overcomes the problem of varying cut-offs between studies and allows the cut-offs to be appropriate to the specific population. Thus, in the normal, overweight, obese, and very obese categories, body mass index ranged from 18.3 to 29.8, 24.6 to 30.0, 29.0 to 40.0, and ≥34.9 to ≥40.0, respectively. Using population specific cut-offs for body mass index is an established practice in other areas of medicine, including using lower body mass index cut-offs for obesity in Asian than white populations since lower cut-offs have been associated with increased risks of cardiovascular disease.102
Future research is needed to try to determine why overweight and obese women are at risk of preterm birth, and to determine effective methods of weight loss in women of childbearing age before pregnancy.
In conclusion, overweight and obese women have higher risks of preterm birth before 32 weeks and induced preterm birth before 37 weeks, and accounting for publication bias, possible preterm birth before 37 weeks overall. Unlike many causes of preterm birth, maternal overweight and obesity represent a potentially preventable cause of the leading source of neonatal mortality and morbidity and morbidity through childhood.7 Surveillance for preterm birth should be considered in overweight and obese women. Moreover, although some of the inductions may have been medically indicated, some were likely not, and represent another area for clinicians to focus on for the prevention of preterm birth. The beneficial effects of maternal overweight or obesity on low birth weight were higher in developing countries than developed countries and disappeared when publication bias was taken into account. Clinicians need to be aware that overweight or obesity in women is not protective against having infants of low birth weight and should consider surveillance when indicated. Ideally, overweight or obese women should have prepregnancy counselling so that they are informed of their perinatal risks and can try to optimise their weight before pregnancy.
Appendix 1: search strategy
Appendix 2: study quality according to six types of bias
Appendix 3: post hoc sensitivity analyses
Appendix 4: funnel plots using trim and fill analysis
We thank Elizabeth Uleryk, chief librarian at The Hospital for Sick Children, Toronto, Canada, for her help in developing the search strategy.
Members of Knowledge Synthesis Group on determinants of preterm birth/low birthweight: Prakesh Shah, associate professor, Department of Paediatrics, Mount Sinai Hospital and Department of Health Policy, Management and Evaluation, University of Toronto, Toronto, Canada; Arne Ohlsson, professor emeritus, Department of Paediatrics, Mount Sinai Hospital and Departments of Paediatrics, Obstetrics and Gynaecology, and Health Policy, Management and Evaluation, University of Toronto, Canada; Vibhuti Shah, associate professor, Department of Paediatrics, Mount Sinai Hospital and Department of Health Policy, Management and Evaluation, University of Toronto, Toronto, Canada; Kellie E Murphy, associate professor, Department of Obstetrics and Gynecology, Mount Sinai Hospital and University of Toronto, Canada; Sarah D McDonald, associate professor, Division of Maternal-Fetal Medicine, Departments of Obstetrics and Gynecology and Diagnostic Imaging, McMaster University, Hamilton, Canada; Eileen Hutton, associate professor, Department of Obstetrics and Gynecology, McMaster University, Hamilton, Canada; Christine Newburn-Cook, associate professor and associate dean of research, Faculty of Nursing, University of Alberta, Edmonton, Canada; Corine Frick, adjunct professor, Faculty of Nursing, University of Calgary, Calgary, Canada; Fran Scott, associate professor, Dalla Lana School of Public Health, University of Toronto and Toronto Public Health, Toronto, Canada; Victoria Allen, associate professor, Department of Obstetrics and Gynaecology, Dalhousie University, Halifax, Canada; and Joseph Beyene, associate professor and John D Cameron endowed chair in genetic epidemiology, McMaster University, Department of Clinical Epidemiology and Biostatistics.
Contributors: All authors conceived and designed the study, analysed and interpreted the data, critically revised the manuscript for important intellectual content, and approved the final versions. SDMcD had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis. She drafted the manuscript and is guarantor.
Competing interests: All authors have completed the unified competing interest form and declare that: (1) this work was supported by a Canadian Institute of Health Research (CIHR) operating grant (No KRS 86242), that SDMcD is supported by a CIHR new investigator award, that ZH was supported by a state scholarship fund by the China Scholarship Council, and that JB is supported by a CIHR grant (No 84392); (2) SDMcD, ZH, SM, and JB have no relationships with any companies that might have an interest in the submitted work in the previous 3 years; (3) their spouses, partners, or children have no financial relationships that may be relevant to the submitted work; and (4) SDMcD, ZH, SM, and JB have no non-financial interests that may be relevant to the submitted work. CIHR and the China Scholarship Council had no role in analyses, writing of the report, interpretation of data or the decision to submit the manuscript.
Ethical approval: Not required.
Data sharing: No additional data available.
Cite this as: BMJ 2010;341:c3428